I have made similar calculations about a year ago and came to the conclusion that Bitcoin could be 51% attacked with about 2-4% of the value of the "market cap" (with some variance because price/difficulty relation is not always the same). I based my calculations on rented hardware (cloud mining) but essentially the cost seems to be similar.
So I agree:
Bitcoin is strong.But Kakmakr's answer reminded me of a scenario I never was able to get out of my mind, where I guess it could be possible for an attacker to obtain large profits with his attack:
If a double spend were done, investors in Bitcoin will question the security of this technology and the validity of the transactions. [...] A massive dump will follow and the price will take a nose dive. An attack of this magnitude will cause panic and the attackers will gain very little in the long run, because the price will keep going down. The cost to sustain such an attack, also outweigh the gains of doing this.
What if the attacker just wanted that -
to crash the price - for example, because he's massively shorting coins and would profit of a crash?
He would have to create a double spend to prove that he attacked the chain, but he can do that with his own addresses. So he would not even face charges of theft. (He could, however, face the charge of "damage to property" as he's interfering in the functioning of the Bitcoin system, or something similar.)
This attack scenario, however, depends massively on the amount of coins which can be shorted. Unfortunately, the new future markets would just allow these kinds of massive short operations.
Anybody able to "rebutt" that scenario?